The first banner ad went live on HotWired (now Wired) in October 1994. It was for AT&T. The agency that made it said 44 percent of the people who saw it clicked.

Nobody has seen that number since. The usual figure for display ads today sits around 0.15 percent. Treat that as a rough industry average. It moves by format and vertical. The banner did not get worse. It got common.

That is the oldest rule in marketing technology. AI is not an exception. A new tool pays the most when it is new and few people have it. After that the payout does not go to the tool. It goes to whoever does the opposite of what the tool made easy.

The advantage was never the tool. It was the gap.

Early email got opened because an inbox had twelve messages in it. Early Facebook pages reached their own followers for free. Early search ads cost cents. In each case the early adopter was not smarter. They were alone. The return came from the gap between what they could do and what everyone else could do. The gap closed the moment the tool became standard.

AI creative ran that whole cycle in about three years. In 2023 an AI-generated image stopped the scroll because you had never seen one. A founder who could ship ten ad variations in an afternoon had a real edge over one who waited two weeks for a designer.

Now every brand in your category has the same models, the same prompts and the same afternoon. Ten variations is the floor. The gap is closed. The tool pays nothing extra. It only keeps you level.

When everyone uses the same tool the same way, the outputs converge. A converged feed has a cost you can measure.

In December 2024, NielsenIQ put more than 2000 people in front of AI-generated ads and measured brain activity for about 150 of them. People identified most of the AI ads on instinct. They described them as annoying, boring and confusing. Even the ads rated as high quality left a weaker memory trace than traditional ads.

Read that last finding again. You are not paying for impressions. You are paying to be remembered. A creative that is cheap to make and weak to remember is not a saving.

In 1982, BBH launched black Levi’s with a poster of a flock of white sheep and one black sheep walking the opposite direction. The line was “When the world zigs, zag.”

The zig today is synthetic. The simplest zag available to you is being human.

(Salman Khan put that on a t-shirt years ago and built a clothing brand out of it. He was early. It turns out Being Human is also a media strategy.)

Some brands have already made the move, and made it loudly. Dove pledged in 2024 never to use AI to represent real women in its ads. In October 2025, Aerie promised never to use AI-generated bodies or people in its marketing, and it makes its creators promise the same. Aerie then reported a 23 percent comparable sales increase for the fourth quarter of 2025 and came out of that quarter up 21 percent in brand awareness.

Be careful with that number. It covers the whole brand, not the pledge alone. No honest analyst would hand all of it to one campaign. The direction is clear enough. In a feed where everything is generated, a real face, a real location and a real mistake are the things that look different. Different is what gets noticed.

If you stop here you will draw the wrong conclusion. You will decide that AI is the problem and a human hand is the fix.

It is not that simple. Coca-Cola has now made its “Holidays Are Coming” Christmas ad with AI two years running. The internet was furious both times. System1, which tests ads on ordinary viewers, gave the 2025 version 5.9 stars. That is the maximum score an ad can get. Most viewers simply saw a warm Christmas ad with trucks in it. In the associations they reported, AI was not mentioned.

Why did Coke get away with it? Because the ad was not built on AI. It was built on thirty years of the same story, the same trucks, the same music. The strategy and the storytelling were already there. AI only did the rendering.

Here is the ranking as I see it:

Strategy and story, made by humans, where human-ness is visible. The strongest position, and the most expensive.

Strategy and story, made with AI. Still works. Coke proved it.

Human-made, with no strategy and no story. A real photo of nothing in particular. Authentic and forgettable.

AI-made, with no strategy and no story. This is slop. This is most of the feed.

Being human is a signal. It is not a strategy. A handmade ad with nothing to say loses to an AI ad with a clear idea, every time. What you cannot generate is the decision about what to say, to whom, and why they should care. That part was always the job.

You do not need to make everything by hand. Spend the human budget where the audience checks for a human:

Faces and bodies. Fashion, beauty, fitness, anything where the buyer asks “would this look like that on me?”

Trust claims. Testimonials, founder messages, anything that says “believe me.”

The first impression. The one creative a new customer meets first.

Keep AI for where nobody checks: resizing, versioning, localization, the fortieth variant of a retargeting banner, the first draft you will rewrite anyway.

One test this month, and it produces a number. Take your best-performing AI creative and build one visibly human version of the same idea: real person, real setting, same message, same offer. Run both to the same audience with equal budget until each has at least 1,000 clicks.

Compare cost per acquisition, not click-through rate. If the human version wins by more than its extra production cost, you have found where your category rewards the zag. If it does not, you have saved yourself an expensive opinion.

The tool was an advantage while it was rare. It is rare no longer. Now the advantage is the direction you walk.